DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined
under the first inventor to file provisions of the AIA .
Response to Arguments
Applicant's arguments filed 2/24/26 have been fully considered but they are not
persuasive.
Regarding applicant Response to Claim Rejections Under 35 U.S.C. § 101.
The applicant argues that the claims are not directed to a judicial exception because they involve real-time processing using specialized classifiers and cannot be practically performed in the human mind.
The examiner respectfully disagreed with the applicant’s argument because the claim recites “determining a label plausibility…based on a tree-based classifier and an image transformation classifier”, which constitutes data analysis using mathematical models. Under current USPTO guidance, a claim recites a mental process when it encompasses acts of evaluation, judgment, or classification of data, even if performed using a computer, unless the claim recites a specific technological implementation that precludes such characterization (see MPEP § 2106.04(a)(2)). The claim here does not recite any specific algorithm structure, training technique, or hardware implementation that would meaningfully limit the classifiers beyond generic functional language. The mere recitation of “tree-based classifier” and “image transformation classifier” does not remove the claim from mental process grouping, as these are broad categories of mathematical models used for classification. Moreover, the assertion that the claimed processing is “real-time” or computationally intensive is not reflected in the claim language and therefore cannot confer eligibility (see MPEP § 2106.04(d)(1)).
The applicant further argues that the claim integrates the alleged exception into a practical application under Step 2A, Prong Two.
The examiner respectfully disagrees with the applicant’s argument because with respect to MPEP § 2106.05(c) (transformation), the recited “modifying a control parameter of a building equipment” does not constitute a meaningful transformation of an article to a different state or thing. The claim does not specify what control parameter is modified, how it is modified, or how such modification changes the physical operation of the equipment in a concrete manner. Instead, the limitation is recited at a high level of abstraction and merely applies the result of the abstract analysis, which is indicative of insignificant post-solution activity (see MPEP § 2106.05(g)).
Regarding MPEP § 2106.05(b) (particular machine), although the claim references a “building automation system” this merely a generic technological environment. The claim does not impose any meaningful limitations on how the system operates or how the classifiers are integrated into the system. The building automation system is therefore not a particular machine integral to the claimed invention, but rather a field of use (see MPEP § 2106.05(h)).
The applicant’s reliance on MPEP § 2106.05(a) (improvement to technology) is likewise not persuasive. While the specification may describe improvements in label accuracy or diagnostics, the claim itself does not recite a specific improvement to the functioning of a computer or to building automation technology. The claim broadly recites determining label plausibility using generic classifiers and modifying a control parameter, without specifying any particular technical mechanism that improves system operation. As such, the claim is directed to the use of generic machine learning techniques to analyze data and apply the results, rather than a specific technological improvement (see MPEP §2106.04(d)(1) and §2106.05(a)).
The applicant further argues that, Under Step 2B, the combination of classifiers operating on the same input and subsequent control action provides an inventive concept.
The examiner respectfully disagrees with the applicant’s argument because the additional elements, both individually and as ordered combination, amount to well-understood, routine, and conventional activities. The use of multiple classifiers (e.g., ensemble or parallel models) to analyze the same dataset is a common practice in machine learning, and applying the result of such analysis to adjust system parameters is likewise conventional in control systems. The claim does not recite any unconventional arrangement, specific interaction between the classifiers that yields a technical improvement, or any non-generic control technique. Instead, the claim follows a conventional sequence of data collection, analysis, and application of results, which does not amount to significantly more than the abstract idea (see MPEP §2106.05(d)).
Finally, the applicant’s reliance on the August 4, 2025 USPTO memorandum regarding A-related claims is not persuasive.
While the memorandum clarifies that certain A_-based claims may fall outside the mental process grouping when they cannot practically be performed in the human mind, this principle applies only when the claim recites specific technological implementations that meaningfully limits the scope. Here, the claim recites the classifiers at high level of generality without technical details, and thus still encompasses abstract data analysis. Accordingly, the claim remains directed to a judicial exception.
For the forgoing reasons, the rejection under 35 U.S.C. § 101 is maintained.
Regarding applicant Response to Claim Rejection Under 35 U.S.C. § 103.
Fundamentally different Technical Problems
The applicant argues that the cited references address fundamentally different problems (sematic tag generation vs, label plausibility validation), and therefore lack motivation to combine.
The examiner respectfully disagrees with the applicant’s argument because it improperly focuses on specific problem statements in isolation rather than the teachings of the references as a whole under a broadest reasonable interpretation (BRI).
While Papadopoulos describes generating semantic data tags, it explicitly teaches applying machine learning models to analyze time-series building data and assign or evaluate labels based on learned patterns (see ¶¶ [0038] and [0111]). Under BRI, such classification inherently involves evaluating whether data corresponds to a given label, which reasonably encompasses assessing label correctness or plausibility. Thus, the distinction between “label generation” and “label validation” is not a structural or functional distinction in the claimed method, rather a difference in intended use or interpretation of the same classification output.
Further, it is well established that a reference is not limited to its stated purpose, and can be relied upon for all that it reasonably teaches to one having ordinary skill in the art. Accordingly, the cited references are analogous in that they address classification and interpretation of building automation time-series data, and the alleged difference in problem statement does not preclude combination.
“”image Transformation Classifier” Operating “Distinctly” from Tree-Based Classifier
The applicant argues that Simmons discloses only a preprocessing pipeline followed by a single classifier, rather than two distinct classifiers operating on the same input.
The examiner respectfully disagrees with the applicant’s argument because Simmons both:
Applying a transformation (e.g., GAF/MTF) to time-series data and classifying the
transformed data using a neural network (see ¶¶ [0123] and [0134]-[0136]), and
Separately extracting statistical features and applying classifiers (e.g., decision
trees and random forests) to those features (see ¶¶ [0091]-[0092] and [0125]-[0126]).
Thus, Simmons explicitly discloses multiple model pathways (e.g., feature-based classifiers and transformation-based neural network classifiers) that process representations derived from the same underlying time-series data. Moreover, Simmons further teaches combining outputs from multiple models (e.g., blender model (see ¶¶ [0140]-[0142]), which reasonably suggests parallel or ensemble operation.
Under BRI, “receiving the same data input” encompasses receiving underlying dataset, even if different representations (raw vs. transformed) are used. Likewise, “operating distinctly” is met by separate model pipelines (e.g., tree-based vs. neural network classifiers) that independently process the data. Therefore, the combination of Papadopoulos and Simmons reasonably teaches or suggests the claimed dual-classifier architecture.
Lack of Motivation to Combine the References
The applicant argues that there is no motivation to combine because Papadopoulos already achieves its purpose and does not require transformation techniques.
The examiner respectfully disagrees with the applicant’s argument because a motivation to combine does not require that a reference be deficient. Rather, it is sufficient that one of ordinary skill in the art would have been motivated to improve accuracy, robustness, or generalization. Here, Simmons explicitly teaches that transforming time-series data into multidimensional representations (e.g., GAF/MTF) improves pattern recognition and classification performance (see ¶¶ [0134]-[0136]).
Accordingly, it would have been obvious to incorporate such transformation-based modeling into the system of Papadopoulos because doing so would improve classification accuracy and robustness across diverse building data sources, thereby enhancing the reliability of label determination. This is a well-recognized design incentive and does not rely on impermissible hindsight.
Missing Disclosure of Key Claim Elements
The applicant argues that the references fail to disclose “label plausibility” as defined in the application.
The examiner respectfully disagrees with the applicant’s argument because the limitation “determining a label plausibility”, under BRI, broadly encompasses evaluating whether a label corresponds to input data. As discussed above, Papadopoulos teaches supervised classification using labeled training data and output probabilities for labels (see ¶¶ [0038] and [0111]), which inherently reflects the likelihood or plausibility of a label given the data. Similarly, Simmons teaches classifiers that output likelihoods for multiple classes (see ¶ [0090]). Such probabilistic outputs reasonably correspond to determining label plausibility.
Applicant’s narrower interpretation based on the specification is not commensurate with the claim language and therefore is not adopted.
Claims 5 and 15 Rejection Over Papadopoulos, Simmons, and Amiri
The applicant argues that Amiri does not cure deficiencies.
The examiner respectfully disagrees with the applicant’s argument because it attacks the references individually rather than the combination as a whole.
Amiri is relied upon for teaching transformation of time series data into image-like representations (e.g., RGB matrices) for classification (see ¶ [0070]). This further reinforces the teaching of Simmons regarding transformation-based classification and supports the use of image-based classifiers operating on time-series-derived representations. The combination of Papadopoulos, Simmons, and Amiri therefore collectively teaches the claimed elements.
No Motivation to Combine Three References
The applicant argues that combining three references constitutes hindsight.
The examiner respectfully disagrees with the applicant’s argument because: each reference contributes a known technique:
Papadopoulos discloses supervised classification of building data
Simmons discloses transformation-based and feature-based classification pipelines.
Amiri discloses image-based representations of time-series data
Combining known techniques to improve classification performance is predictable use of prior art elements according to their established functions, which is proper under KSR. The rationale is based on improving accuracy and robustness, not hindsight reconstruction.
Applicant’s arguments with respect to amended claims, filed 2/24/26, have been
considered but are moot.
Amended claims 1 and 11 now includes the newly added limitations: “modifying a control parameter of a building equipment based on the determined label plausibility” (claim 1) and “wherein the processor modifies a control parameter of a building equipment based on the determined label plausibility” (claim 11). Examiner agrees the combination of Papadopoulos, Simmons is silent on these features. However, a new reference M
u
¨
ller ( NPL: “Identifying Mislabeled Instances in Classification Datasets”) has been found to remedy the above-noted deficiencies of Papadopoulos, Simmons. Accordingly, the rejection is set forth below in this Office action.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1-20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to a judicial exception (abstract idea) without significantly more.
Under Step 1 of the 2019 Revised Patent Subject Matter Eligibility Guidance, the claims are directed to a process (claim 1, a method) or a machine (claim 11, an system), which are statutory categories.
However, evaluating claim 1, under Step 2A, Prong One, the claim is
directed to the judicial exception of an abstract idea using the grouping of a mathematical relationship/mental process. The limitations include:
determining a label plausibility for each set of timeseries data and the corresponding label associated with the set of timeseries data based on a tree-based classifier and an image transformation classifier, the tree-based classifier and the image transformation classifier receiving the same data input and operating distinctly from each other; and modifying a control parameter of a building equipment based on the determined label plausibility.
These limitations constitute a mathematical concept and mental process, i.e., analyzing data using mathematical models to generate probabilities or confidence measures. Such classifier-based probability determination is an abstract idea involving mathematical relationship and statistical modeling.
Next, Step 2A, Prong Two evaluates whether additional elements of the claim “integrate the abstract idea into a practical application” in a manner that imposes a meaningful limit on the judicial exception, such that the claim is more than a drafting effort designed to monopolize the exception. The claim does not recite additional elements that integrate the judicial exception into a practical application.
The recited “building automation system” merely serves as a source of data, and the claim does not recite controlling building equipment, modifying system operation, improving sensor functionality , or otherwise effecting a technological improvement. The classifiers are applied to data in a generic manner to evaluate label plausibility, which amounts to mere data analysis.
Therefore, the claims are directed to an abstract idea.
At Step 2B, consideration is given to additional elements that may make the abstract idea significantly more. Under Step 2B, there are no additional elements that make the claim significantly more than the abstract idea.
The additional elements of “receiving timeseries data from the building automation system” and “the image transformation classifier receiving the same data input and operating distinctly from each other” are considered insignificant extra-solution activity of collecting data that is not sufficient to integrate the claim into a particular practical application. The act of data gathering by the sensors is considered insufficient to elevate the claim to a practical application.
The additional element “modifying a control parameter of a building equipment based on the determined label plausibility” is recited at a high level of generality and merely applies the result of the abstract data analysis to a generic technological environment without specifying how the control parameter is determined, what specific control action is performed, or how such modification improves operation of the building automation system, thus, it amounts to insignificant post-solution activity (see MPEP § 2106.05(g)) and mere instructions to apply the exception (see MPEP § 2106.05(f)). The claim does not recite any specific improvement to computer functionality or to building control technology, but instead uses generic classifiers (e.g., tree-based and neural network/image transformation classifiers) as tools to perform the abstract analysis.
Accordingly, the claim is directed to an abstract idea and fails to recite “significantly more”, and it considered not eligible for patent protection under 35 U.S.C. § 101.
Dependent claims 2-10 do not add anything which would render the claimed invention a patent eligible application of the abstract idea. The claims merely extend (or narrow) the abstract idea which do not amount for "significant more" because they merely add details to the algorithm which forms the abstract idea as discussed above.
The additional element “training each of the tree-based classifier and the image transformation classifier to identify a pattern associated with each set of timeseries data and the corresponding label, for a particular data type, wherein training each classifier includes inputting positive examples and negative examples to the classifier, the positive examples including data points of the same label and the negative examples including data points of different labels” (claim 10) is considered performing mathematical calculation which falls within the “mathematical concept” grouping of abstract ideas (see Example 47, in the 2024 Guidance Update on Patent Subject Matter Eligibility, Including on Artificial Intelligence).
Claim 11 is rejected 35 USC § 101 for the same rationale as in claim 1.
This judicial exception is not integrated into a practical application because the remaining elements amount to no more than general purpose computer components programmed to perform the abstract ideas. As set forth in the 2019 Eligibility Guidance, 84 Fed. Reg. at 55 “merely include[ing] instructions to implement an abstract idea on a computer” is an example of when an abstract idea has not been integrated into a practical application
Dependent claims 12-20 do not add anything which would render the claimed invention a patent eligible application of the abstract idea. The claims merely extend (or narrow) the abstract idea which do not amount for "significant more" because they merely add details to the algorithm which forms the abstract idea as discussed above.
This judicial exception is not integrated into a practical application because the remaining elements amount to no more than general purpose computer components programmed to perform the abstract ideas. As set forth in the 2019 Eligibility Guidance, 84 Fed. Reg. at 55 “merely include[ing] instructions to implement an abstract idea on a computer” is an example of when an abstract idea has not been integrated into a practical application
The additional element “trains each of the tree-based classifier and the image transformation classifier to identify a pattern associated with each set of timeseries data and the corresponding label, for a particular data type, wherein trains each classifier includes inputting positive examples and negative examples to the classifier, the positive examples including data points of the same label and the negative examples including data points of different labels” (claim 20) is considered performing mathematical calculation which falls within the “mathematical concept” grouping of abstract ideas (see Example 47, in the 2024 Guidance Update on Patent Subject Matter Eligibility, Including on Artificial Intelligence).
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 1-4, 7, 9-14, 17, 19 and 20 are rejected under 35 U.S.C. 103 as being
unpatentable over Papadopoulos (Pub. No. US 2021/0109485) in view of Simmons et al. (Pub. No. US 2023/0304690) (hereinafter Simmons) and further in view of M
u
¨
ller et al. (NPL: “Identifying Mislabeled Instances in Classification Datasets”, IEEE 2019) (hereinafter M
u
¨
ller).
As per claims 1, 7, 11 and 17, Papadopoulos teaches receiving timeseries data from the building automation system and determining a label plausibility for each set of timeseries data and the corresponding label associated with the set of timeseries data based on a tree-based classifier (see ¶¶ [0038]-[0040], i.e., receiving time series building data from a building automation/control network and applying multiple machine learning models, including a random forest, to evaluate the time series and generate semantic data tags based on model outputs and probability thresholds. In particular ¶ [0038] discloses that the systems obtain probability distributions of semantic data tags from each model and generate semantic data tags if the probability exceeds a threshold, which teaches determining a probability-based assessment of how well a label fits the time-series data, i.e. a measure of plausibility. Thus, Papadopoulos teaches determining label plausibility for time series data based on tree-based classifier (random forest) and other machine learning models operating on the same data input).
However, Papadopoulos fails to explicitly teach reprocessing the time series using
an image transformation classifier, the tree-based classifier and the image transformation classifier receiving the same data input and operating distinctly from each other (emphasis underlined).
Simmons, however, teaches applying multidimensional transformation processes, including GAF (e.g., GASF/GADF) and Markov Transition Fields (MTF), to time series data to generate transformed multidimensional matrices and providing the transformed matrices as input to a neural network such as a conventional neural network to obtain classification outputs (see ¶¶ [0092]-[0093], [0134]-[0136] and [0141]). A multidimensional matrix representation processed by a conventional neural network reasonably reads, under broadest reasonable interpretation, on an “image transformation classifier”. It would have been obvious to one having ordinary skill in the art before the effective filling date of the claimed invention to incorporate the GAF/MTF-based multidimensional transformation and CNN classifier of Simmons into the multi-model probability framework of Papadopoulos because Simmons that such transformations generate multidimensional representations that encode correlations between segments of the time series data and transition probabilities, thereby improving classification accuracy and enhancing the reliability of probability-based label plausibility determinations in building automation systems. Accordingly, the combination teaches determining a label plausibility for each set of time series data and its associated label based on both a tree-based classifier and an image transformation classifier receiving the same input and operating distinctly from each other.
However, the combination of Papadopoulos and Simmons fails to explicitly teach modifying a control parameter of a building equipment based on the determined label plausibility.
M
u
¨
ller, however, teaches evaluating the correctness of labels in a dataset by comparing classifier predictions and actual labels to identify instances where labels are likely incorrect or inconsistent with the learned data distribution, thereby determining a likelihood that a given label is erroneous (i.e., label plausibility) (see pages 3-4, sections III.- IV.). M
u
¨
ller further teaches that such detection of mislabeled data instances is used to trigger corrective actions, such as filtering, relabeling, or otherwise modifying subsequent processing based on the determined likelihood of label correctness (see pages 2-3, section II. “Related Work” ). It would have been obvious to one having ordinary skill in the art before the effective filling date of the claimed invention to incorporate the label evaluation techniques of M
u
¨
ller into the combination of Papadopoulos and Simmons to determine a plausibility of labels associated with time-series data because machine learning systems operating on building automation data rely on accurate labels for reliable classification, thereby enabling identification of inconsistent or implausible labels and improving the accuracy of downstream fault detection and diagnostics. Furthermore, Simmons teaches that outputs of data analysis may be used to detect faults and trigger actions related to building equipment operation (e.g., identifying equipment needing service or adjustment) (see ¶ [0106]). Accordingly, it would have been obvious to use the determined label plausibility to modify a control parameter of the building equipment based on the reliability of the classification results because improving the reliability of classification outcomes enhances the effectiveness of control and diagnostic decisions, thereby yielding a more robust fault detection and building control system as claimed.
As per claims 2 and 12, the combination of Papadopoulos and Simmons teaches the system as stated above. Papadopoulos fails to explicitly teach parsing the timeseries data into a plurality of predetermined time periods, the plurality of predetermined time periods being associated with a calendar period type; extracting a plurality of statistical features for the timeseries data; and identifying a data type associated with one or more field devices of the building automation system.
However, Simmons further teaches parsing time-series data into predetermined time windows associated with calendar-based periods, such daytime and nighttime windows (see ¶ [0098]), thereby meeting the limitation of parsing the time-series data into a plurality or predetermined time periods associated with a calendar period type. Simmons also teaches extracting a plurality of statistical features from the time-series data, including mean, median, standard deviation, minimum, maximum, percentile values, autocorrelation, Fourier harmonics, correlation values, and ratios of consecutive constant values (see ¶¶ [0091] and [0129]-[0133]), which satisfies the limitation of extracting a plurality of statistical features. Additionally, Simmons identifying point types and equipment types associated with points/devices in a building control network, including sensors, actuators, setpoints, alarms, air handling units, chillers, and other building field devices (see ¶¶ [0103]-[0105] and [0118]-[0122]), which reasonably reads on identifying a data type associated with one or more field devices of the building automation system. It would have been obvious to one having ordinary skill in the art before the effective filling date of the claimed invention to implement the calendar-based windowing, statistical feature extraction, and point/equipment type identification techniques of Simmons within the labeling framework of Papadopoulos because Simmons teaches that time-window segmentation and statistical feature extraction allow the statistical model to distinguish between different point types and equipment types based on their statistical and temporal characteristics (see ¶¶ [0091], [0098] and [0133]), thereby improving the accuracy of label determinations in building automation systems.
As per claims 3 and 13, the combination of Papadopoulos and Simmons teaches the system as stated above. Papadopoulos further teaches that determining the label plausibility comprises: generating, based on the tree-based classifier (i.e., Random Forest), a probability corresponding to an association between each set of timeseries data and the corresponding label (see ¶ [0038], which discloses that the systems obtain probability distributions of semantic data tags from each model and generate semantic data tags based on those probabilities, including generating a tag if the probability exceeds a threshold).
As per claims 4 and 14, the combination of Papadopoulos and Simmons teaches the system as stated above. Papadopoulos further teaches that the tree-based classifier is a random forest classifier (see ¶ [0038]); and generating the probability includes determining that a particular label of the data point is wrong or exhibits abnormal behavior based on the probability failing to exceed a predetermined threshold (see ¶ [0038], which discloses that generating a probability corresponding to the association between the time series data and a label, and determining whether to accept or reject the label based on whether the probability exceeds a predetermined threshold. Under broadest reasonable interpretation, generating the label only when the probability exceeds the threshold reasonably reads on determining that a particular label does not properly correspond to the time series data when the probability fails to exceed the threshold).
As per claims 9 and 19, the combination of Papadopoulos and Simmons teaches the system as stated above. Papadopoulos further teaches modifying one or more labels based on the label plausibility (see ¶ [0111], i.e., modifying semantic data tags (i.e., labels) based on probability outputs of the neural network, wherein the neural network generates a probability distribution showing the probability that different semantic data tags are correct outputs and ¶¶ [0038]-[0040], i.e., generates sematic data tags when the probability exceeds a threshold and applies rules to probability outputs) (The examiner notes that selecting or reassigning semantic data tags based on probability distributions and threshold comparisons reasonably reads on modifying one or more labels based on label plausibility as claimed).
As per claims 10 and 20, the combination of Papadopoulos and Simmons teaches the system as stated above.
Papadopoulos teaches applying machine learning models, including a random forest classifier, to building automation system time-series data to generate semantic data tags (see ¶ [0038]) and further teaches training such models using historical labeled data stored in a building management system database (see ¶ [0039]). In particular, Papadopoulos discloses that the neural networks are trained by providing labeled training datasets including data points tagged with semantic data outputs, comparing outputs to correct semantic data tags, and modifying model parameters via back propagation to improve classification accuracy (see ¶ [0111]). (The examiner notes that such supervised training involves providing training examples associated with particular labels and adjusting model parameters based on differences between predicted and correct labels for relevant data types).
Simmons further teaches applying a multidimensional transformation process (e.g., Gramian Angular Field or Markov Transition Field) to time-series data to obtain transformed multidimensional matrices (see ¶¶ [0134-[0136]), and providing the transformed data as input to neural network classifier, including a conventional neural network (see ¶ [0123]), to determine labeled point types and equipment types based on outputs representing likelihoods associated with multiple classes (see ¶ [0090]), Simmons further teaches training the statistical model using training data corresponding to multiple labeled point types and equipment types (see ¶ [0120]) which likewise involves supervised learning using labeled training data corresponding to different classes. It would have been obvious to one having ordinary skill in the art before the effective filling date of the claimed invention to train both the tree-based classifier of Papadopoulos and the image-transformation neural network classifier of Simmons using labeled examples corresponding to particular labels and other labels for each type because supervised classification techniques conventionally rely on labeled training data representing multiple classes in order to learn discriminative patterns, thereby enabling each classifier to identify patterns associated with particular time-series data and their corresponding labels as claimed.
Claims 5 and 15 are rejected under 35 U.S.C. 103 as being unpatentable over
Papadopoulos in view of Simmons and further in view of Amiri et al. (Pub. No. US 2022/0407597) (hereinafter Amiri).
As per claims 5 and 15, the combination of Papadopoulos and Simmons teaches the system as stated above except that determining the label plausibility comprises: generating, based on the image transformation classifier, a probability corresponding to an association between each set of timeseries data and the corresponding label, generating the probability including transforming each set of timeseries data to an RGB image.
However, Amiri teaches converting multi-variate time-series datasets into matrices (see ¶¶ [0054]-[0056]), stacking the matrices into a larger multi-variate matrix including a resultant image having an RGB format (see ¶ [0070]), and passing the resultant RGB image to a three-channel neural network such as a CNN for classification and anomaly detection (see ¶¶ [0060]-[0061] and [0070]). It would have been obvious to one having ordinary skill in the art before the effective filling date of the claimed invention to modify the teaching of the combination of Papadopoulos and Simmons to utilize the time-series to RGB image transformation technique of Amiri, because Amiri explicitly teaches that transforming muti-variate time-series data into RGB image representations enables the use of image-based neural network classifiers for time-series classification, thereby providing an alternative, known input representation that enables image-based neural network classification of time-series data.
Examiner’s Notes
Claims 6, 8, 16 and 18 distinguish over the prior art.
Please note that claim 18 is objected to for minor informalities as stated
above.
Regarding claims 6 and 16, none of the prior art of record either singularly or in
combination anticipates or render obvious a method or a system for fault detection and diagnostics of a building automation system comprising: applying at least one neural network to the transformed RGB images, the at least one neural network including a CNN and a multi-layer perceptron; and combining at least one statistical features with each transformed RGB image, in combination with the rest of the claim limitations as claimed and defined by the applicant.
Regarding claims 8 and 18, none of the prior art of record either singularly or in
combination anticipates or render obvious a method or a system for fault detection and diagnostics of a building automation system comprising: wherein determining the label plausibility comprises: determining the label plausibility based on a first probability of the tree-based classifier and a second probability of the image transformation classifier, the label plausibility including a percentage or ratio representing how well each label associates with the timeseries data corresponding to the label, in combination with the rest of the claim limitations as claimed and defined by the applicant.
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in
this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Contact information
Any inquiry concerning this communication or earlier communications from the
examiner should be directed to MOHAMED CHARIOUI whose telephone number is (571)272-2213. The examiner can normally be reached Monday through Friday, from 9 am to 6 pm.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Andrew Schechter can be reached on (571) 272-2302. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
Information regarding the status of an application may be obtained from the Patent Application Information Retrieval (PAIR) system. Status information for published applications may be obtained from either Private PAIR or Public PAIR. Status information for unpublished applications is available through Private PAIR only. For more information about the PAIR system, see http://pair-direct.uspto.gov. Should you have questions on access to the Private PAIR system, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free).
Mohamed Charioui
/MOHAMED CHARIOUI/Primary Examiner, Art Unit 2857
Prosecution Insights